Transition RANS to LES during simulation
 

Hi everyone,

I am looking for a way to switch from RANS to LES during a simulation run.
I want to investigate slug flow in a horizontal pipe and especially the secondary flow within a plug.
The pipe is 10m long so I cannot do the whole simulation with LES due to limited resources.

Is it possible to switch from RANS turbulence model to LES for the pipe at a certain time-step close to the end?

Or is possible to simulate the last half meter of the pipe with LES and the rest of the pipe with RANS?

Since I am quite a newbie regarding OpenFOAM I would be very thankful for any clues.

Regards,
Lorenz

RANS and LES usually have much different requirements when it comes to the mesh. LES needing more cells, so you should have two different meshes set up.

one thing you can do is to run the simulation in RANS on the first mesh and use the results (from any time directory) as the input for your LES simulation. you should use the mapFields utility. http://www.openfoam.com/docs/user/st...-utilities.php

You can also try some hybrid approach, such as the DES, though I don't think you'll save a lot of computational resources comparing with pure LES.

V.

Thanks for the quick reply.
The mapFields approach sounds good.
Is it possible map the conditions at every cross sectional area of the pipe (in axial direction)?
I am asking this because it would be perfect for me if I could map the conditions right before the formation of a slug near the end of the pipe.
The DES approach is not beneficial for me as I just need LES to investigate the formation of a slug and its flow through the pipe.

The DES approach resolves only the boundary layer with a RANS formulation, while the core of the flow exhibits a consistent LES behavior, thus probably it would be a good solution for your case. The problem (in my opinion) is that the RANS solution of the b.l. still requires a very fine grid (though probably slightly coarser than the pure LES requirements) so probably you will not gain that much reduction in computational time.
The mapping option is interesting, I think you should think about two options: 1) resolve the first part of the domain separately with a RANS model, and then map only the outlet values as inlet BC for a LES involving only the second part of the domain (I think that there are some proper mapping functions to do so, but I don't know exactly what they are); do the same, but resolving also the second part of the domain with RANS modeling and then map the whole converged second-part solution on your LES mesh, thus initializing not only the inlet BC's but also the internal domain.

Regards

V.